Sanjay Chandlekar working under the supervision of Dr. Sujit Gujar presented his research work on A Novel Demand Response Model and Method for Peak Reduction in Smart Grids — PowerTAC  at the 22nd International Conference on Autonomous Agents and Multiagent Systems (AAMAS’23) held in London from 29 May to 2 June and his other research work on  Multi-armed Bandit Based Tariff Generation Strategy for Multi-Agent Smart Grid Systems at the 11th International Workshop on Engineering Multiagent Systems (EMAS’23) held from 29 – 30 May at London.

•  A Novel Demand Response Model and Method for Peak Reduction in Smart Grids — PowerTAC

 Authors  Sanjay Chandlekar, Arthik Boroju, Shweta Jain and Sujit Gujar explain their research work here:

 We study the Demand Response behavior of smart grid customers in response to the offered discounts for peak reduction. We propose a model that depicts the probability of a customer reducing its load as a function of the discounts offered. This function is parametrized by the rate of reduction (RR). We provide an optimal algorithm, MJS–ExpResponse, that allocates the discounts to each customer by maximizing the expected reduction under a budget constraint. When RRs are unknown, we propose a Multi-Armed Bandit based online algorithm, namely MJSUCB–ExpResponse, to learn RRs. We experimentally show that it exhibits sublinear regret and showcase its efficacy in a real-world smart grid system using the PowerTAC simulator as a test bed.

• Multi-armed Bandit Based Tariff Generation Strategy for Multi-Agent Smart Grid Systems

 Authors Sanjay Chandlekar, Easwar Subramanian and Sujit Gujar explain their research work here:

 The emergence of smart grid technology has opened the door for wide-scale automation in decision-making. A distribution company, an integral part of a smart grid system, has to procure electricity from the wholesale market and then sell it to customers in the retail market by publishing attractive tariff contracts. It can deploy autonomous agents to make decisions on its behalf. In this work, we describe the tariff contracts generation strategy of one such autonomous agent, which is based on a Contextual Multi-armed Bandit (ConMAB) based learning technique to generate tariff contracts for various types of customers in the retail market of smart grids. We particularly utilize the Exponential-weight algorithm for Exploration and Exploitation (EXP-3) for ConMAB-based learning. We call our proposed strategy GenerateTariffs-EXP3. Our previous work shows that maintaining an appropriate market share in the retail market yields high net revenue. Thus, we first present a game-theoretic analysis that determines an optimal level of market share. Then we train our proposed strategy to achieve and maintain the suggested level of market share by adapting to the market situation and revising the tariff contracts periodically. We validate our proposed strategy in PowerTAC, a close-to real-world smart grid simulator, and showcase that it is able to maintain the suggested market share.